Tuhinga 15: 27–41 Copyright © Te Papa Museum of New Zealand (2004)

Geographic variation in macrantha (): morphology and flavonoid chemistry

Michael J. Bayly,1 Alison V. Kellow,1 Rebecca Ansell,1 Kevin A. Mitchell,2 and Kenneth R. Markham2 (1) Museum of New Zealand Te Papa Tongarewa, PO Box 467, Wellington, New Zealand [email protected] [email protected] (2) Industrial Research Limited, PO Box 31 310, Lower Hutt, New Zealand [email protected] [email protected]

ABSTRACT: This paper investigates geographic variation in Hebe macrantha, endemic to mountains of South Island, New Zealand. It assesses the attributes, distribution, and appropriate taxonomic status of its two previously described varieties, the validity and circumscriptions of which have been questioned by some taxonomists. Morphometric analyses support the existence of two distinguishable entities that equate with previous circumscriptions of these varieties, whose continued recognition is recommended. Variety brachyphylla occurs in the north of the species’ geographic range, from the Anatoki Range, northwest Nelson, to the Hanmer Range, north Canterbury. Variety macrantha is geo- graphically more widespread and morphologically more variable. It occurs in the south of the species’ range, south and west from Mt Haast (southwest Nelson) and Lewis Pass (south Nelson/north Canterbury), as well as at Lake Tennyson (southeast Nelson), the only locality at which specimens assigned to both varieties occur. The two varieties are morphologically most similar at localities close to their geographic interface. Patterns of variation in flavonoids neither strongly support nor contradict recognition of the two varieties. Analysed samples share a similar and, within Hebe, distinctive flavonoid profile, and the distribution of two flavonoid glycosides is partly correlated with the morpholog- ical circumscription of varieties (the correlation is incomplete in samples taken near the geographic interface of varieties). A distribution map and a table of key morphological differences between varieties are provided.

KEYWORDS: Hebe macrantha, Plantaginaceae, flavonoids, New Zealand flora. 28 Tuhinga, Number 15 (2004)

Fig. 1 Some morphological features of Hebe macrantha: (A) habit of a of var. brachyphylla on Mt Arthur; (B) sprigs of var. macran- tha (left) and var. brachyphylla; (C) top view of shoot apex, showing that young diverge early (i.e. there is no prominent ‘lead bud’ like that seen in many hebes); (D) lower and upper leaf surfaces of var. macrantha (top) and var. brachyphylla; (E) showing frontal view of ; (F) lateral view of flower; (G) inflorescence (left) and infructescence; (H) septicidal (left) and loculicidal views of capsule (voucher specimens: A, from same population as WELT 80740; B (left), D (upper), WELT 82429; B (right), E, F, G, WELT 83544; C, D (lower), WELT 82554; H, WELT 83545). Variation in Hebe macrantha 29

Introduction Kellow et al. 2003a, b). It aims to identify and describe patterns of geographic variation within H. macrantha Hebe macrantha (Hook.f.) Cockayne et Allan is a distinc- using data on morphology and leaf flavonoids, and to re- tive sub- from alpine regions of South Island, New assess its infraspecific classification. In particular, it seeks Zealand. It is readily distinguishable from other members to determine whether or not there are grounds for the of Hebe by its much larger (see Figs. 1E and 1F), recognition of two taxa, and if so, to identify clearly their fruits that are laterally compressed (with the septum at distinguishing features, map their distributions, and assess right angles to the widest diameter; see Fig. 1H), and leaves their appropriate taxonomic rank. that are prominently petiolate (see Fig. 1D), have toothed This study was initiated primarily because an initial margins, and diverge early in bud (see Fig. 1C). Moore (in survey of specimens suggested that the leaf dimensions Allan 1961), in an informal infrageneric classification of specified by Moore (in Allan 1961) were insufficient to Hebe, placed H. macrantha in the monotypic grouping identify reliably all specimens to variety rank. A further “Grandiflorae”, this name referring to the characteristically motivation was the opinion of Prof. P. Garnock-Jones large flowers. Alternatively, Heads (1987), considered H. (Victoria University of Wellington), based primarily on macrantha best removed from Hebe and placed in field observations of morphological variation in the area Parahebe, whose members it resembles in many morpho- around Lake Tennyson, south Nelson Province (personal logical features. More recent analyses of nuclear ribosomal communication 1997), that there were probably insuffi- DNA sequences (Wagstaff & Garnock-Jones 1998, cient grounds for recognition of two taxa, and that the Wagstaff et al. 2002) group H. macrantha with other classification of H. macrantha warranted thorough assess- members of Hebe, sister to all other species of the ment. Heads (1994a) also suggested that ‘the two varieties (as defined by Wagstaff et al. 2002, and followed here). of the species and their distribution require revision’. Cheeseman (1906) reported that specimens of Hebe macrantha (as macrantha) ‘from Mount Arthur and other parts of the Nelson District have shorter, broad- Materials and methods er leaves, more numerous and smaller flowers than Morphometric analysis is usual in Canterbury and Otago …’, and provided the variety name brachyphylla to distinguish these collections Fifty-one herbarium specimens from WELT and CHR from more typical forms. Cockayne and Allan (1926) (see Appendix 1; herbarium abbreviations follow agreed with the presence of two varieties of H. macrantha Holmgren et al. (1990)), covering most of the geographic (using the name var. vera for the typical form), and added range of Hebe macrantha, were scored for 16 morphologi- that var. brachyphylla ‘occurs not only in the North-western cal characters (see Table 1). The geographic distribution of Botanical District, as given in the Manual, but also in the these specimens is shown in Fig. 4, although three speci- wetter part of the North-eastern District’. Moore (in Allan mens – WELT 13110 (‘western Amuri District’, presum- 1961) also recognised two varieties in Hebe macrantha, and ably somewhere to the west of Hanmer Springs), WELT further defined them, giving leaf dimensions of c. 1.5–2.5 13101 (‘Ashburton mountains’), and WELT 13111 x 0.5–1.0 cm for var. macrantha and c. 1.3–1.5 x 0.9– (‘Otago’) – have only vague locality information and could 1.0 cm for var. brachyphylla. This classification, including not be mapped. Unless each of the fragments on an two varieties, has been followed in subsequent, general herbarium specimen was known to be from a single indi- works on the New Zealand flora (e.g. Eagle 1982, Druce vidual (e.g. in the case of MJB collections), or fragments 1993, Wilson & Galloway 1993, Mark & Adams 1995). were very small or were uniform in appearance, only one Distribution maps for the two varieties are given by Heads fragment was scored per specimen. In one case, WELT (1994a) and Macdonald (1982; with the two maps labelled 82420, two fragments on the one sheet were known to be the wrong way around), with the former providing no data from separate , and each was treated as a separate on morphological circumscription, and the latter largely specimen (specimens 12a and 12b in Appendix 1). repeating details given by Moore (in Allan 1961). Vegetative characters were measured for each of 10 leaves The study presented here is part of ongoing work and 10 internodes per specimen. Characters of flowers and toward a revised classification of Hebe (e.g. Bayly et al. were measured for up to eight flowers or 2000, 2001, 2002, 2003; Garnock-Jones et al. 2000; inflorescences per specimen (where available). Character 30 Tuhinga, Number 15 (2004)

Table 1 Characters used in the morphometric analysis. to the Kolmogorov-Smirnov test for normality (Sokal & Rohlf 1969) and Levene’s median test for homogeneity of variance (Levene 1960). For characters that passed tests for Internode length (INT) (mm) normality and equal variance (P > 0.05), t-tests were used Lamina length (LL) (mm) to compare groups. For characters that failed either test, the Lamina width (LW) (mm) Mann-Whitney Rank Sum Test was used to compare Lamina length/lamina width (LL/LW) groups. These analyses were performed using SigmaStat for Lamina length/total leaf length (LL/TLL) Windows Version 2.0. Distance from leaf base (including petiole) to widest point (DWP) (mm) Examination of additional Distance to widest point/total leaf length herbarium specimens (including petiole) (DWP/TLL) Results of the morphometric analysis were assessed by Petiole length (PET) (mm) reference to further herbarium material. All specimens of Number of teeth on one side of leaf (the bigger Hebe macrantha at WELT and CHR (95 additional speci- number if uneven) mens) were assessed for the five most discriminating mor- Peduncle length (mm) phological characters, and on the basis of their geographic Rachis length (mm) origin, to determine whether variation among all speci- Rachis length/total inflorescence length mens was congruent with, and adequately reflected by, Number of flowers per inflorescence that shown in the morphometric study. Length of lowermost bracts on inflorescences (mm) Pedicel length (mm) Analysis of leaf flavonoids Calyx lobe length (mm) Leaf samples collected from wild populations were placed in press-seal plastic bags with silica gel and couriered to the laboratories of Industrial Research, Lower Hutt. Voucher specimens are listed in Table 3, and sampling localities are means were calculated for each specimen and used as input indicated on Fig. 4. for multivariate and univariate analyses. Flavonoids were extracted from dried samples, and For multivariate analyses, data for each character were were separated and identified using the methods of two- range-standardised, such that the range of all characters dimensional paper chromatography (2DPC) and high- equalled one unit, giving all characters equal weight performance liquid chromatography outlined by Bayly et (recommended by Milligan & Cooper 1988). Pairwise al. (2002, 2003). For the purpose of this report, results are distances between specimens were calculated using the shown only for compounds definitely present in two or Manhattan metric, and the distance matrix was subjected more samples. A number of additional compounds were to clustering by the unweighted pair-group method using found in only one sample, or could not be distinguished arithmetic averages (UPGMA; Sneath & Sokal 1973) and with certainty (in a range of samples). Details of these to ordination using non-metric multidimensional scaling compounds, which do not influence conclusions present- (NMDS; Kruskal 1964a, b) in two dimensions, both ed here, will be published elsewhere in a more implemented by NTSYSpc Version 2.11. The results of comprehensive summary of Hebe flavonoids (Markham et NMDS are dependent on starting conditions, being based al. in preparation). on iterations from an initial configuration of points, and are sensitive to local minima. Here, the procedure was repeated at least 30 times with random starting conditions, Results and the best result taken as that with the lowest stress value Morphometric analysis – i.e. best fit to the distance matrix. UPGMA analysis produced a single dendrogram, i.e. there Univariate comparisons were made between the two were no ‘ties’ between groups at any stage in the tree- major groups of specimens identified by UPGMA analysis. building process. The primary dichotomy on this tree Prior to comparison, data for each character were subjected distinguishes two clusters, groups A and B (see Fig. 2), Variation in Hebe macrantha 31

Fig. 2 UPGMA tree derived from analysis of morphometric characters. Specimen numbers are as given in Appendix 1. In the classification adopted here: Group A = Hebe macrantha var. brachyphylla; Group B = H. macrantha var. macrantha.

which are also separated in the NMDS ordination (see Variation in individual morphological characters with- Fig. 3). Specimens of groups A and B overlap in their in and between the two groups, together with the results of geographic distribution only slightly (see Fig. 4), and cor- univariate statistical comparisons, are summarised in Fig. 5. respond broadly with previously accepted geographic All characters show some overlap in range between the limits of var. brachyphylla and var. macrantha, respectively. two groups, but most comparisons, nonetheless, show Group A includes specimens from localities between the statistically significant differences between groups (the only Peel Range, northwest Nelson, and Mt St Patrick and Mt characters not significantly differing between the two Percival, northeast Canterbury. Group B includes all spec- groups are lamina width (see Fig. 5C), the ratio of lamina imens from localities south and west of Mt Haast and length to total leaf length (see Fig. 5E), and rachis length Lewis Pass, as well as one specimen from Lake Tennyson. (see Fig. 5K)). The characters providing greatest differenti- Lake Tennyson is the only locality at which specimens of ation between groups (the least amount of overlap, in terms both groups occur. It is evident from both the UPGMA of mean values for specimens) are the length of lowermost dendrogram and the NMDS ordination that specimens of inflorescence bracts (see Fig. 5N), the number of teeth on Group A, which is geographically more restricted, are gen- leaf margins (see Fig. 5I), the ratio of leaf lamina length to erally more similar to each other than are specimens of the width (see Fig. 5D), the distance from leaf bases to their more widespread, and more variable, Group B. widest points (DWP; see Fig. 5F), and peduncle length 32 Tuhinga, Number 15 (2004)

Fig. 3 Two-dimensional NMDS ordination derived from analysis of morphometric characters. Specimen numbers are as given in Appendix 1. Groups A and B are the same as shown in Fig. 2. In the classification adopted here: Group A = Hebe macrantha var. brachyphylla; Group B = H. macrantha var. macrantha.

(see Fig. 5J). In general, specimens of Group B frequently geographic interface between these groups. In particular, also have longer leaf laminae (see Fig. 5B), more obovate three Group B specimens – one from Mt Haast (specimen leaves (i.e. with a greater ratio of DWP to total leaf length; 24), one from Lake Tennyson (specimen 23), and one from see Fig. 5G), longer petioles (see Fig. 5H), a lower ratio of ‘western Amuri district’ (specimen 22) – are placed in rachis length to total inflorescence length (see Fig. 5L), positions intermediate between Group A and remaining longer pedicels (see Fig. 5O), and longer calyx lobes (see Group B specimens. Likewise, among the Group A speci- Fig. 5P). Group B specimens sometimes also have longer mens closest to those of Group B are one from Mt Mantell internodes (see Fig. 5A) and fewer flowers per inflorescence (specimen 14), one from Lake Tennyson (specimen 16), (see Fig. 5M), but there is substantial overlap between and one from Baldy (specimen 15); these three specimens, groups in these characters (the probability, or p value, that together with one other from Lake Tennyson, specimen 17, group means/medians are different is between 0.05 and also form a group in the UPGMA dendrogram. 0.001). Within Group A, patterns of similarity among speci- Results of the NMDS show that specimens of Group mens, apart from the similarity of those specimens from A and Group B are most similar at localities close to the near the southwest of the group’s geographic range Variation in Hebe macrantha 33

Fig. 4 Map of South Island, New Zealand, showing the distribution of Hebe macrantha and some examples of leaf shape and size. Closed symbols indicate localities represented by specimens in the morphometric analysis. Names are given only for localities mentioned in the text, and those from which flavonoid samples were collected. Numbers under leaf outlines are WELT numbers for the specimens from which they were drawn. Scale bar is for leaf outlines. 34 Tuhinga, Number 15 (2004)

Fig. 5 Graphs showing variation in 16 morphological characters. Each data point represents one specimen (of those listed in Appendix 1), and shows the mean, maximum, and minimum measurement for the character. Specimens are sorted into the two groups (Groups A and B) recovered by UPGMA analysis (see Fig. 2), and arranged within these groups by mean value for each character. Asterisks at the bottom right of graphs indicate statistically significant differences between groups for that character (* = p < 0.05; ** = p < 0.001), based on a Mann-Whitney Rank Sum Test (asterisks in parentheses), or t-test (no parentheses). Units on graphs are millimetres, except for I and M (which are absolute numbers), and ratios given on D, E, G, and L (which are unitless). Character abbreviations are as given in Table 1. In the classification adopted here: Group A = Hebe macrantha var. brachyphylla; Group B = H. macrantha var. macrantha. Variation in Hebe macrantha 35

Table 2 Ranges of variation, based on all herbarium specimens at WELT and CHR, for the five morphological characters best discriminating Groups A and B.

Character Group A Group B (var. brachyphylla)(var. macrantha)

Lamina length/lamina width (0.97–)1.1–2.2(–2.9) 1.2–3.3(–3.8) Distance from leaf base (including petiole) to widest point (mm) (4.4–)6–11(–13.8) (5.1–)10–20(–23.2) Number of teeth on one side of leaf (0–)1–4(–5)* (2–)3–7(–11) Peduncle length (mm) (1.5–)3.0–13.0 (6–)7.0–30.7 Length of lowermost bracts on inflorescence (mm) 2–4(–8)† (4–)5–9.1

* Value = 0 on only a few leaves of WELT 80740 (Mt Arthur). † Extreme value of 8 mm seen only on CHR 333979 (Mt Arthur), which has unusually large flowers. Lowermost bracts on all other specimens ≤ 5 mm.

(specimens 14, 16, and 15), show no strong geographic required only minor modification of the ranges of some association. In contrast, within Group B, there is some morphological characters for each group. The morpholog- additional, though partial, geographic correlation in pat- ical ranges of Group A and Group B for the five most dis- terns of similarity among specimens. For instance, speci- criminating characters (using all specimens at WELT and mens from Amuri Pass (specimens 26 and 27) are placed CHR) are shown in Table 2 (compare with Fig. 5 for minor near that from near Lewis Pass (specimen 25) in the differences from character ranges in the subset of speci- NMDS ordination. Together, these specimens are close to mens used in morphometric analyses). those from the Hope (specimen 28) and Poulter valleys (specimen 30), and all of these are close to those from Leaf flavonoid composition Browning Pass (specimen 33), Arthurs Pass (specimens 31 Flavonoids identified in Hebe macrantha in general resem- and 32), and the Griffin Range (specimen 29). Also, five of ble those found in other Hebe species. Flavonoids with the the six Fiordland specimens (specimens 43, 44, 45, 47, basic apigenin and luteolin oxygenation patterns, and and 48) cluster closely in both the NMDS ordination and glycosylated at the 7-hydroxyl group (e.g. compounds the UPGMA dendrogram. There is, therefore, a general 4 and 10), are found throughout H. macrantha, as also is trend in the arrangement of specimens on axis I of the luteolin glycosylated at the 4-hydroxyl (1a). Luteolin ordination that correlates roughly, though not absolutely, glycosylated at the 3-hydroxyl (1b), however, is only con- with latitude – i.e. south to north, moving from left to spicuously evident in the Group A samples and in one right on the axis. Group B sample (WELT 81716). Modification of the basic apigenin and luteolin aglycone types in H. macrantha Examination of additional includes C-glycosylation (in some Group B specimens herbarium specimens only) and 6- and 8-oxygenation (in both groups). Further Variation in additional herbarium specimens at WELT and biosynthetic elaboration of the 6-hydroxyluteolin CHR was generally consistent with that seen in the mor- flavonoid nucleus, as evidenced by the accumulation of phometric study. Specimens that would, on the basis of a 6-methoxyluteolin glycoside (3), is seen consistently only geographic origin, be placed in either Group A or Group B in Group B. Also confined to Group B, although not were generally also, in terms of the five most discriminating in all samples, is the unidentified flavonoid ‘r’, which morphological features, morphologically consistent with possesses an absorption spectrum that suggests it is an these groups. Placing all specimens in groups in this way 8-hydroxyluteolin-8-O-glycoside. 36 Tuhinga, Number 15 (2004)

Table 3 Presence/absence of flavonoid compounds in each of the samples analysed. Key: +, conspicuous presence on 2DPC; w, weak presence on 2DPC; ?, possible or uncertain occurrence (after a structure name it indicates a tentative identification); L, luteolin; A, apigenin; OMe, methoxy; OH, hydroxy. Each compound is given a number/letter code that is part of a database (being developed at Industrial Research) of all major flavonoid compounds in Hebe. Asterisks (*) after voucher numbers indicate specimens also included in morphometric analysis.

10 3e 5e 24 4 12a 1b 1a 8b 9 3 5a r -xyloglucoside] -glucoglucoside] O O -glucoside -rhamnoglucoside] -[2- -glucoside -[2- O -glucoside O O O O C [2- -glucoside -glucoside

Voucher Locality -glucoside -glucoside O O O (WELT no.) O O- nidentified flavonoid 8OHL-7- 6OHA(like)-glycoside 6OHA(like)-glycoside A-6,8-di- U L-7- A-7- L-7- 6OHL-7- 6OMeL-7- L-4’- 6OHL-7- L-3’-

Group A (var. brachyphylla) 80653§ Peel Range + + + +++++ + 80653§ Peel Range + + +++++ + 82420* Mt Murchison w + + w + + + + 80980* Lake Tennyson + + + + + + + + +

Group B (var. macrantha) 81671* Mt Haast + w w + + + w + + w 81716* Amuri Pass +++++ +++ ++w 80507 Sealy Range w + w + ? + + w + + 81622 Sealy Range + + + ? w + + + + 82429 Near Mt Brewster + ? + + + + + + w 82430 Near Mt Brewster + + + + ? + + + + 80866* Gertrude Saddle + + + + w +++++ 80867* Gertrude Saddle + + + + ? + + w + + w

§ These are samples from two individuals in the same population (first row = M.J. Bayly 472; second row = M.J. Bayly 474). Variation in Hebe macrantha 37

These observations indicate some partial differences morphological and geographic separation of the groups, in the biosynthetic enzyme activities between the two and the presence of somewhat intermediate specimens. groups. For example, luteolin-3’-O-glucosyltransferase This choice also allows continued use of the existing variety (which produces compound 1b from luteolin in Group A) names, var. macrantha and var. brachyphylla, avoiding the is absent or largely ineffective in Group B. Conversely, creation of new names or combinations. 6-hydroxyluteolin-6-O-methyltransferase (which converts Defining the morphological limits of var. macrantha 6-hydroxyluteolin-7-O-glucoside into 6-methoxyluteolin- and var. brachyphylla is not straightforward, given that 7-O-glucoside, 3) is chiefly functional only in Group B some specimens are somewhat intermediate between the plants. It is noteworthy that both of these differences two main groups. Here, the two varieties are defined such between the two groups are least pronounced in samples that they correspond to the two groups identified in the from near their geographic interface, i.e. the only Group A UPGMA dendrogram, and have the key characteristics sample producing compound 3 was from Lake Tennyson, outlined in Table 2. The taxonomic placement of some while the only Group B sample producing a conspicuous intermediate specimens could be debated, and other, amount of 1b was from Amuri Pass. perhaps less mathematical methods, could possibly be used A distinctive feature in the flavonoid constituents of to define groups; if anything, the pattern of morphological Hebe macrantha (both groups) is the occurrence of com- variation might allow a slightly broader definition of var. pounds 3e and 5e. The aglycone moiety is the same for brachyphylla, but not of var. macrantha. However the two each compound and, although unidentified, is similar to, varieties are separated, the fact would remain that most but not identical with, 6-hydroxyapigenin, based on the specimens could readily be placed under one variety or the uv/visible absorption spectra. Glycosides of this aglycone other, but that a small number would be more problematic. are not found in quantities above trace levels in other For problematic specimens, understanding where they hebes, except in the informal group “Semiflagriformes” of stand in relation to the pattern of variation in the species is Moore (in Allan 1961), which could, on phylogenetic probably more important than what name they are given. grounds (Wagstaff et al. 2002), be considered a distinct Further collecting could, of course, extend both the genus, Leonohebe Heads (but in a narrower circumscrip- geographic and morphological ranges of the two varieties, tion than proposed by Heads 1987, 1994b, c). and, particularly in south Nelson, could potentially affect interpretations of the boundaries between them. Patterns of variation in leaf flavonoids neither strong- Discussion ly support nor contradict recognition of the two varieties. Appropriate classification The similarity of the flavonoid profiles of all samples sup- Analyses of morphological data differentiate two groups of ports the view that members of the species are closely specimens, with a small number of specimens (e.g. speci- related. Partial correlation in the distribution of both com- mens 22, 23, and 24 of Group B) having somewhat inter- pounds 3 and 1b with the morphological circumscriptions mediate characteristics. These groups have substantial, only of varieties is interesting, but further sampling would narrowly sympatric, geographic ranges, indicating that be required to judge the consistency and significance of patterns of morphological variation are far from random, these results. Lack of substantial flavonoid variation with- and supporting the view that the groups are worthy of in species, including between infraspecific taxa, is not taxonomic recognition. Given that all specimens share unusual (Bohm 1987), and is also seen in other Hebe distinctive characters of flowers, fruits, and leaves (see species, e.g. H. stenophylla (Mitchell et al. 2001) and ‘Introduction’), have similar flavonoid profiles, and that H. stricta (unpublished data). variation is mostly quantitative, with the ranges of all characters overlapping in the two morphological groups, Distribution of varieties recognition at infraspecific rank is most appropriate. The The distribution map presented here (see Fig. 4) shows ranks of subspecies or variety could possibly equally be some differences from those previously presented by used (given the different ways in which they are used by Macdonald (1982) and Heads (1994a), particularly in different authors), but here we suggest the continued use southern Nelson (map of Macdonald) and in the northern of variety rank. This is chiefly because of the degree of part of the range of var. macrantha (map of Heads). 38 Tuhinga, Number 15 (2004)

Differences partly reflect different sources of herbarium in glacial periods to separated refugia (say one in Nelson specimens (Macdonald used CHR and CANU; Heads and one further south), and separated populations differen- used a range of herbarium specimens, as well as published tiated from each other and then subsequently expanded and unpublished records, and personal field observations) their ranges with the coming of interglacial conditions. and collecting subsequent to publication of the earlier The data provided here cannot distinguish between these maps. They also reflect different interpretations of the or other (possibly more complex) hypotheses. The partial identity of some specimens, e.g. Heads (1994a) identified geographic correlation of variation, especially in var. specimen(s) from Mt Haast as var. brachyphylla, and macrantha, is possibly suggestive of a north–south cline of that/those from Mt Terako as var. macrantha (for which variation, but these patterns are not clear-cut, and variation reverse identifications are accepted here). within var. brachyphylla is less strongly correlated with Wilson (1991) provided a map of Hebe macrantha, latitude. Use of further sources of data (e.g. DNA sequenc- not distinguishing or mentioning the different varieties, in ing or fingerprinting) could provide further insight into the the provinces of Canterbury and Westland. That map historical basis of observed variation. shows a similar distribution to that presented here. However, it was based not only on herbarium specimens, but also on published and unpublished reports and exten- Acknowledgements sive field observation, and shows more detail of the species David Glenny, Debra Wotton, Phil Garnock-Jones, Bill distribution in some areas. Malcolm, Barbara Brown, Ilse Breitwieser, and Rainer Vogt all provided assistance with fieldwork. Bill Malcolm Historical interpretation provided the photographs for Figs 1C, 1D, 1E, 1F, and Several explanations could be offered for the pattern of 1H. Patrick Brownsey, Phil Garnock-Jones, and Leon variation in Hebe macrantha, i.e. that there are two more Perrie provided constructive comments on the draft or less well-defined groups, with some geographic overlap manuscript. The Director and staff of CHR provided and some morphologically intermediate specimens. This access to collections and a loan of herbarium specimens. pattern could result from ongoing, in situ, differentiation The Department of Conservation provided permits to of forms from within a cline of variation. It could also collect in reserves under their control. This research was result from the introgression of two previously differen- funded by a grant from the Foundation for Research, tiated forms, e.g. if the species’ distribution was restricted Science and Technology (contract MNZX0201). Variation in Hebe macrantha 39

Holmgren, P.K., Holmgren, N.H., and Barnett, L.C. (1990). References Index Herbariorum. Part I: The Herbaria of the World. 8th Allan, H.H. (1961). Flora of New Zealand. Vol. I. Wellington: edn. Bronx, New York: New York Botanical Garden, Government Printer. pp. i–liv + 1085 pp. International Association for Plant . pp. i–x + Bayly, M.J., Garnock-Jones, P.J., Brownsey, P.J., Mitchell, 693 pp. K.A., and Markham, K.R. (2000). A taxonomic revision Kellow, A.V., Bayly, M.J., Mitchell, K.A., Markham, K.R., of the Hebe parviflora complex (Scrophulariaceae), based and Garnock-Jones, P.J. (2003a). Variation in morphol- on morphology and flavonoid chemistry. New Zealand ogy and flavonoid chemistry in Journal of Botany 38: 165–90. (Scrophulariaceae), including a revised subspecific classi- Bayly, M.J., Garnock-Jones, P.J., Mitchell, K.A., Markham, fication. New Zealand Journal of Botany 41: 233–53. K.R., and Brownsey, P.J. (2001). Description and Kellow, A.V., Bayly, M.J., Mitchell, K.A., Markham, K.R, flavonoid chemistry of Hebe calcicola (Scrophulariaceae), and Brownsey, P.J. (2003b). A taxonomic revision of a new species from north-west Nelson, New Zealand. Hebe informal group “Connatae” (Plantaginaceae), based New Zealand Journal of Botany 39: 55–67. on morphology and flavonoid chemistry. New Zealand Bayly, M.J., Kellow, A.V., Mitchell, K.A., Markham, K.R., Journal of Botany 41: 613–35. de Lange, P.J., Harper, G.E., Garnock-Jones, P.J., and Kruskal, J.B. (1964a). Multidimensional scaling by optimiz- Brownsey, P.J. (2002). Descriptions and flavonoid chem- ing goodness of fit to a nonmetric hypothesis. istry of new taxa in Hebe sect. Subdistichae (Scrophu- Psychometrika 29: 1–27. lariaceae). New Zealand Journal of Botany 40: 571–602. Kruskal, J.B. (1964b). Nonmetric multidimensional scaling: Bayly, M.J., Kellow, A.V., de Lange, P.J., Mitchell, K., a numerical method. Psychometrika 29: 28–42. Markham, K.R., Garnock-Jones, P.J., and Brownsey, P.J. Levene, H. (1960). Robust tests for equality of variances. In: (2003). Geographic variation in morphology and Olkin, I. (ed.) Contributions to Probability and Statistics. flavonoid chemistry in Hebe pubescens and H. bollonsii Stanford: Stanford University Press. pp. 278–92. (Scrophulariaceae), including a new infraspecific classifi- Macdonald, A.D. (1982). Distribution maps of Hebe in cation for H. pubescens. New Zealand Journal of Botany Canterbury. Canterbury Botanical Society Journal 16: 41: 23–53. 73–6. Bohm, B.A. (1987). Intraspecific flavonoid variation. Mark, A.F. and Adams, N.M. (1995). New Zealand Alpine Botanical Review 53: 197–279. Plants. Auckland: Godwit. 269 pp. Cheeseman, T.F. (1906). Manual of the New Zealand Flora. Milligan, G.W. and Cooper, M.C. (1988). A study of stan- 1st edn. Wellington: Government Printer. pp. i–xxxvi + dardisation of variables in cluster analyses. Journal of 1199 pp. Classification 5: 181–204. Cockayne, L. and Allan, H.H. (1926). The present taxonomic Mitchell, K.A., Markham, K.R., and Bayly, M.J. (2001). status of the New Zealand species of Hebe. Transactions Flavonoid characters contributing to the taxonomic and Proceedings of the New Zealand Institute 57: 11–47. revision of the Hebe parviflora complex. Phytochemistry Druce, A.P. (1993). Indigenous vascular plants of New 56: 453–61. Zealand (9th revision). Unpublished checklist held at Sneath, P.H.A. and Sokal, R.R. (1973). Numerical Taxonomy. Landcare Research, Lincoln, New Zealand. San Francisco: Freeman. 573 pp. Eagle, A. (1982). Eagle’s Trees and of New Zealand. 2nd Sokal, R.R. and Rohlf, F.J. (1969). Biometry. 1st edn. San series. Auckland: William Collins Publishers Ltd. 382 pp. Francisco: W.H. Freeman and Company. 776 pp. Garnock-Jones, P.J., Bayly, M.J., Lee, W.G., and Rance, B.D. Wagstaff, S.J. and Garnock-Jones, P.J. (1998). Evolution and (2000). Hebe arganthera (Scrophulariaceae), a new biogeography of the Hebe complex (Scrophulariaceae) species from calcareous outcrops in Fiordland, New inferred from ITS sequences. New Zealand Journal of Zealand. New Zealand Journal of Botany 38: 379–88. Botany 36: 425–37. Heads, M.J. (1987). New names in New Zealand Scrophu- Wagstaff, S.J., Bayly, M.J., Garnock-Jones, P.J., and Albach, lariaceae. Botanical Society of Otago Newsletter 5: 4–11. D.C. (2002). Classification, origin, and diversification of Heads, M.J. (1994a). A biogeographic review of Parahebe the New Zealand hebes (Scrophulariaceae). Annals of the (Scrophulariaceae). Botanical Journal of the Linnean Missouri Botanical Garden 89: 38–63. Society 115: 65–89. Wilson, H. (1991). Distribution maps of small-leaved shrubs Heads, M.J. (1994b). Biogeography and evolution in the in Canterbury and Westland. Canterbury Botanical Hebe complex (Scrophulariaceae): Leonohebe and Society Journal 25: 3–81. Chionohebe. Candollea 49: 81–119. Wilson, H. and Galloway, T. (1993). Small-leaved Shrubs of Heads, M.J. (1994c). Morphology, architecture and tax- New Zealand. Christchurch: Manuka Press. 305 pp. onomy in the Hebe complex (Scrophulariaceae). Bulletin du Muséum National d’Histoire Naturelle, Section B, Adansonia 16: 163–91. 40 Tuhinga, Number 15 (2004) ppendix 1: details of specimens used in morphometric analysis. A Asterisks (*) indicate whether specimens were scored for vegetative (Veg.) and/or inflorescence/flower (Infl.) characters. (Infl.) and/or inflorescence/flower (Veg.) for vegetative scored Asterisks (*) indicate whether specimens were from Myttons Hut from Myttons Hut Myttons from Lake Peel Mt Luna ridge below corner of lake Malcolm W. WELT 80671WELT 80652Thoms Creek Range, just east of WELT Peel Range, along track to Lake Peel Peel 41˚ 8'S 80654WELT 41˚ 9'S Range, along track to Peel 80740WELT 35'E 172˚ 13108WELT Along track to Mt Arthur & D. Glenny Kellow 37'E 490, A.V. M.J. Bayly 172˚ 13122WELT Arthur Range, Mt Arthur 07.01.1997 Kellow & A.V. D. Glenny 471, M.J. Bayly CHR 358531 Arthur Range, Mt Arthur 41˚ 9'S 07.01.1997 14905WELT * Ridge Garibaldi * CHR 219913 Spur Arthur Range, Cowin * 37'E 172˚ 41˚ 13'S River, * Wangapeka Branch North 79831 41˚ 13'SWELT Kellow & A.V. 473, D. Glenny M.J. Bayly 42'E 172˚ 41˚ 13'SCHR 355250 07.01.1997 Richmond Range, Mt Fishtail 41'E 172˚ & I. Breitwieser 560, D. Glenny M.J. Bayly 41˚ 24'S range Matiri Nelson, west North 41˚ 18'S 41'E 172˚ 13.02.1997 Cheeseman * T.F. 27'E 172˚ Cheeseman T.F. 18'E 172˚ * 41˚ 14'S 41˚ 27'SCHR 401820 71/103 B.H. Macmillan 41˚ 38'S * Oliver W.R.B. CHR 280430 25'E [Baldy] of Mt Baldy South-south-west 172˚ 30'E 173˚ 17'E 172˚ 80980WELT Tennyson Lake Druce A.P. 42˚ 2'S J. Hadfield Druce near south-west A.P. Tennyson, Lake 24'E 172˚ CHR 228691 01.1886 Druce 23.01.1971 42˚ 13'S A.P. Patrick St Mt 01.1886 * 45'E 172˚ * * & Garnock-Jones 24.01.1956 763, P.J. M.J. Bayly * * 42˚ 12'S * * 18.12.1997 44'E 172˚ * 01.03.1980 11.1932 03.1979 451 Garnock-Jones P.J. * * 42˚ 27'S * * * 01.03.1984 44'E 172˚ * * — * 18.01.1976 * 01.1972 * brachyphylla CSpec. Herb. No. Herb. CSpec. No. LocalityVar. 1 Latitude Longitude Collector(s) Date Veg. Infl. 2 3 4 5 6 7 8 9 10 11 12a&b 82420 WELT 13 Range, Mt Murchison Braeburn 14 CHR 19813215 CHR 50683916 Lakes, Mt Misery Nelson 41˚ 45'S17 Mt Mantell Saddle, of Maruia North 31'E 172˚ 18 42˚ 0'S Kellow19 1467 & A.V. M.J. Bayly CHR 10743 41˚ 56'S20 17'E 172˚ 21Valley Ada Mountains, Spencer 94/41 B.H. Macmillan 40'E 172˚ CHR 333978 13102 WELT 12.01.2001 — Range, Mt Percival Hanmer Range, Mt Percival Hanmer 42˚ 19'S * 36'E 172˚ 42˚ 28'S R.M. Laing 30.01.1994 42˚ 28'S 56'E 172˚ * 56'E 172˚ Wall A. * D. Petrie — 14.01.1949 * * * * 01.1919 10.02.1914 * * * Variation in Hebe macrantha 41 02.1891 * * d .J. Bayly 600, D. Glenny & B. Brown 600, D. Glenny .J. Bayly 28.02.1997 * * .J. Bayly & D. Wotton & D. .J. Bayly Kellow 1287 & A.V. .J. Bayly Kellow 1288 & A.V. .J. Bayly 12.12.1999 30.12.1999 30.12.1999 * * * * & B. Brown 592, D. Glenny .J. Bayly & B. Brown 593, D. Glenny .J. Bayly 25.02.1997 25.02.1997 * * * * . Mackay 20.01.1924 * * .R.B. Oliver.R.B. Oliver.R.B. Oliver 26.01.1928 26.01.1928 * * * 22.01.1950 * * oy .F. Cheeseman.F. — * * .J. Garnock-Jones 18.01.1976 * * & K.A. Ford Buxton R.P. Wardle, . 27.04.1993 * * D.R. Given 11216 & P. Douglass & P. 11216 D.R. Given 03.1978 * A.P. DruceA.P. L. Cockayne 01.1978K. Thomas * Wooff W. & A.D. Campbell, D. Prouting 02.1978 214 Druce A.P. * 22.01.1900 * *D. Petrie * R.A. Wilson — 02.1991 * *D. Petrie * * L. Cockayne MurdochJ.W. 02.1911 04.1920 * * * * 01.1892 — 1912 * * * * * * EP E EM EM E EW EW E EP E E E E—E E E — ** E EW EM ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' EM EW ET EB E EM EM ' ' ' ' ' ' ' 72˚ 43 72˚ 5 72˚ 22 72˚ 45 72˚ 45 72˚ 10 71˚ 17 71˚ 49 71˚ 35 71˚ 35 71˚ 21 70˚ 35 70˚ 36 70˚ 40 70˚ 9 70˚ 6 70˚ 5 70˚ 5 68˚ 44 68˚ 11 68˚ 1 68˚ 1 67˚ 46 67˚ 46 67˚ 55 69˚ 27 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' op Hope Hut vicinity Hut op Hope south of Taramakau river Taramakau south of T macrantha 30 13112 WELT Hill Green River, Poulter Near 42˚ 52 Var. 2223 13110 WELT Amuri District CHR 280428 Western Tennyson Lake above Hillside 42˚ 12 — — Morrison W.G. — * * 44 80866 WELT Gertrude Saddle 44˚ 45 2425 81671 WELT 26 CHR 323544 northeast Mt Haast, slopes27 81715 WELT of Lewis Pass West 28 81716 WELT Amuri Pass CHR 479927 Amuri Pass 42˚ 18 Valley, Hope Upper 3132 42˚ 23 13116 WELT 33 14906 WELT Blimit Arthurs Pass, 34 13099 42˚ 35 WELT Blimit Arthurs Pass, 42˚ 31 35 CHR 499955 Pass divide, Browning Main 42˚ 31 36 CHR 317876Wilberg Mt 37 CHR 469139 south of fork River, Barlow 38 14907 StreamThumb Range, Eric WELT Two 39 42˚ 57 42˚ 55 13123 Alex Knob WELT Josef, Franz 43˚ 18 42˚ 55 13104 WELT Valley Mt Cook, Hooker 43˚ 24 41 Mt Cook, near old Hermitage42 13106 WELT 43 13119 WELT River Hunter Lake Hawea, 43˚ 12 43˚ 26 80874 43˚ 44 WELT Mt Aspiring 43˚ 42 45 near south edge of lake Lake Harris, 46 80867 WELT 47 44˚ 44 44˚ 16 13103 WELT Gertrude Saddle48 13113 WELT Clinton Saddle49 13109 WELT Clinton Saddle50 13101 WELT Valley Clinton 13111 WELT Ashburton mountains 44˚ 23 Otago 44˚ 45 44˚ 48 44˚ 48 — 44˚ 54 — — Potts T.H. — — — * * — * * 40 13105 WELT Range Mt Cook, Sealy 43˚ 45 29 CHR 63421 Range, [Griffin] Griffen 42˚ 50